Cell placement alteration apparatus for integrated circuit chip physical design automation system

• US 5,793,644 A
• Filed: 09/17/1996
• Issued: 08/11/1998
• Est. Priority Date: 04/19/1994
• Status: Expired due to Term

• Current Assignee: Bell Semiconductor, LLC (Hilco Inc. (d/b/a Hilco Global))
• Original Assignee: LSI Logic Corporation (Broadcom, Inc.)
• Inventors: Scepanovic, Ranko, Koford, James S., Boyle, Douglas B., Rostoker, Michael D., Jones, Edwin R.
• Primary Examiner(s): Teska, Kevin J.
• Assistant Examiner(s): PHAN, THAI Q
• Application Number:
  US08/724,025
• Time in Patent Office: 693 Days
• Field of Search: 364/489, 364/490, 364/491, 364/468, 364/513, 395/500, 395/13
• US Class Current: 716/123
• CPC Class Codes: G06F 30/392 Floor-planning or layout, e.g. partitioning or placement

Abstract

A large number of possible placements of cells on an integrated circuit chip are generated and evaluated to determine the placement with the highest fitness. Cells for transposition or "swapping" within each placement using genetic algorithms are selected using greedy algorithms based on the fitness of each cell. The cell fitnesses are evaluated in terms of interconnect congestion, total net wire length or other criteria. Cells are selected for genetic crossover by sorting the cells in order of fitness and multiplying the cell fitnesses by weighting factors that increase non-linearly with rank. The cells are selected using linear random number generation such that cells with higher fitnesses have a higher probability of selection. Cells having lowest fitness are selected for mutation, and transposed to random locations, to adjacent locations, with cells having second worst fitness, to the center of mass of the respective interconnect nets, or with two or more cells in a cyclical manner.

Claims

1. A method for selecting cells for transposition such that a probability of each cell being selected is a predetermined function of a fitness associated with each cell respectively, comprising the steps of:

sorting and ranking said cells in increasing order of fitness;

multiplying said fitnesses by weighting factors that increase non-linearly with cell rank to produce weighted fitnesses respectively, wherein said predetermined function of said fitness is such that said probability of said each cell being selected is substantially linearly proportional to said weighted fitness of said each cell respectively;

computing a weighted fitness summation for said each cell as being substantially equal to the sum of said weighted fitness of said each cell and said weighted fitnesses of said cells having lower fitnesses than said each cell respectively;

generating a random number having a maximum value that is substantially equal to a maximum weighted fitness summation for a first cell placement; and

selecting a cell for transposition which has a weighted fitness summation that is substantially equal to said random number.

2. A method as in claim 1, in which th e sorting and ranking step comprises computing said weighted fitness for said each cell as substantially V_i =F_i ξ

ⁱ, where i is a rank of said each cell in increasing order of fitness, ξ

is a predetermined value, V_i is said weighted fitness, F_i is said fitness and ξ

_i said weighting factor of said each cell respectively.

3. A method as in claim 1, wherein cell selection comprises selecting first and second cells in said first cell placement and a third cell placement respectively for transposition.

4. A method as in claim 3, in which cell selection comprises selecting said first cell as having a lowest fitness among said cells in said first cell placement.

5. A method as in claim 3, in which cell selection comprises selecting said first cell as having a highest fitness among said cells in said first cell placement.

6. A method as in claim 3, in which cell selection comprises selecting said first cell in accordance with a predetermined fitness criteria and selecting a second cell as having a same location in a second cell placement as said first cell has in said first cell placement.

7. A method as in claim 1, in which cell selection comprises selecting said first cell and a second cell in said first cell placement respectively for transposition.

8. A method as in claim 7, in which cell selection comprises selecting said first cell as having a lowest fitness among said cells in said first cell placement.

9. A method as in claim 7, in which cell selection comprises selecting said first cell as having a highest fitness among said cells in said first cell placement.

10. A method as in claim 7, in which cell selection comprises selecting said first cell in accordance with a predetermined fitness criteria and selecting said second cell randomly.

11. A method as in claim 7, in which cell selection comprises selecting said first cell in accordance with a predetermined fitness criteria and selecting said second cell as being adjacent to said first cell.

12. A method as in claim 7, in which cell selection comprises selecting said first cell as having a lowest fitness and said second cell as having a second lowest fitness among said cells in said first cell placement.

13. A method as in claim 7, in which cell selection comprises selecting said first cell in accordance with a predetermined fitness criteria and selecting said second cell as being at substantially a center of mass of an interconnection net to which said first cell is connected.

14. A method as in claim 13, in which cell selection comprises selecting said first cell as having a lowest fitness among said cells of the first cell placement, selecting said second cell as having a lowest fitness among cells that are adjacent to said first cell, and selecting said third cell as having a lowest fitness among cells other than said first cell that are adjacent to said second cell.

15. A method as in claim 1, in which cell selection comprises selecting first, second and third cells in said first cell placement respectively for transposition.

16. A method as in claim 15, in which cell selection comprises selecting said first, second and third cells as having lowest, second lowest and third lowest fitnesses in said first cell placement respectively.

17. A selection processor as in claim 1, wherein the selection processor selects first, second and third cells in said first cell placement respectively for transportation.

18. A selection processor as in claim 17, in which the selection processor selects said first, second and third cells as having lowest, second lowest and third lowest fitnesses in said first cell placement respectively.

19. A selection processor for use with a cell placement alteration apparatus for an integrated circuit chip physical design automation system, comprising:

means for selecting cells such that a probability of each cell being selected is a predetermined function of a fitness of said each cell respectively, wherein the selection processor sorts and ranks said cells in increasing order of fitness and multiplies said fitnesses by weighting factors that increase nonlinearly with cell rank to produce weighted fitnesses respectively, and further wherein said predetermined function of said fitness is such that said probability of said each cell being selected is substantially linearly proportional to said weighted fitness of said each cell respectively;

means for computing a weighted fitness summation for said each cell as being substantially equal to the sum of said weighted fitness of said each cell and said weighted fitnesses of said cells having lower fitnesses than said each cell respectively;

a random number generator for generating a random number having a maximum value that is substantially equal to a maximum weighted fitness summation for a first cell placement; and

a selector for selecting a cell for transposition which has a weighted fitness summation that is substantially equal to said random number.

20. An selection processor as in claim 19, wherein the selection processor computes said weighted fitness for said each cell as substantially V_i =F_i ξ

ⁱ, where i is a rank of said each cell in increasing order of fitness, ξ

is a predetermined value, V_i is said weighted fitness, F_i is said fitness and ξ

_i are said weighting factors of said each cell respectively.

21. A selection processor as in claim 19, wherein the selection processor selects first and second cells in said first cell placement and a third cell placement respectively for transposition.

22. A selection processor as in claim 21, wherein the selection processor selects said first cell as having a lowest fitness among said cells in said first cell placement.

23. A selection processor as in claim 21, wherein the selection processor selects said first cell as having a highest fitness among said cells in said first cell placement.

24. A selection processor as in claim 21, wherein the selection processor selects said first cell in accordance with a predetermined fitness criteria and selects said second cell as having a same location in said second cell placement as said first cell has in said first cell placement.

25. A selection processor as in claim 19, wherein the selection processor selects first and second cells in said first cell placement respectively for transposition.

26. A selection processor as in claim 25, wherein which the selection processor selects said first cell as having a lowest fitness among said cells in said first cell placement.

27. A selection processor as in claim 25, in which the selection processor selects said first cell as having a highest fitness among said cells in said first cell placement.

28. A selection processor as in claim 25, in which the selection processor selects said first cell in accordance with a predetermined fitness criteria and selects said second cell randomly.

29. A selection processor as in claim 25, wherein the selection processor selects said first cell in accordance with a predetermined fitness criteria and selects said second cell as being adjacent to said first cell.

30. A selection processor as in claim 25, wherein the selection processor selects said first cell as having a lowest fitness and said second cell as having a second lowest fitness among said cells in said first cell placement.

31. A selection processor as in claim 25, wherein the selection processor selects said first cell in accordance with a predetermined fitness criteria and selects said second cell as being at substantially a center of mass of an interconnection net to which said first cell is connected.

32. A selection processor as in claim 19, in which said fitnesses of said cells correspond to interconnect congestions of said cells respectively.